Yidu Tech founder and chairperson Gong Yingying will deliver the opening keynote at the HIMSS26 APAC Conference in August, arguing that healthcare AI cannot scale on technical capability alone. Without trust designed into governance structures, clinical validation processes, and accountability mechanisms, she says, AI tools will stall at the pilot stage rather than transforming care delivery.
"The question is not simply whether an AI model is powerful. The real question is whether it can be safely embedded into clinical workflows, whether its reasoning can be traced back to evidence, and whether clinicians can understand when to use it, when to question it, and when to override it," Gong said in an interview with Healthcare IT News.
Her keynote, "The Trust Crisis in AI: The Silent Barrier to Healthcare Transformation," addresses what she calls a prerequisite for scaling - trust that is engineered from the start, not bolted on after deployment.
AI as assistant, not replacement
Gong rejects the framing of AI as a substitute for physicians. Yidu Tech builds its systems on three pillars: high-quality data, clinical-grade evidence, and human-machine collaboration. The AI serves as an intelligent assistant that reduces repetitive work, surfaces relevant evidence faster, and supports more informed decisions.
"The trust crisis often appears when AI is introduced as a 'black box' or as a standalone technology detached from workflow, governance and accountability," she said. Integration into clinical systems must come with clear boundaries, measurable value, and continuous feedback from medical professionals.
Designing accountability into the system
On the question of liability when AI tools make mistakes - a concern highlighted in a recent HIMSS AI Market Study - Gong said accountability must be addressed "by design, not after deployment." Yidu Tech's approach links AI outputs to structured medical knowledge, guidelines, and real-world clinical evidence so clinicians can trace the reasoning behind every suggestion.
Each application carries a defined role in the workflow, whether for information retrieval, risk stratification, or clinical decision support, with different levels of autonomy and human oversight mapped to each. A closed-loop model covers pre-deployment validation, post-deployment monitoring, expert review mechanisms, and continuous improvement. "In healthcare, responsibility cannot sit with a model alone; it must be shared across technology providers, healthcare institutions, clinical governance teams, and regulatory frameworks," Gong said.
Catching mistrust before it hardens
Early signs of eroding clinician confidence are often subtle. Usage drops off. Every output gets manually double-checked, even for low-risk tasks. Users say the tool is "not relevant to my patients" or misaligned with workflow. Sometimes the most telling signal is silence - clinicians stop engaging without lodging complaints.
Organizations should respond by building structured feedback channels, reviewing where trust is weakening, and distinguishing whether the root cause is a model problem, a data problem, a design flaw, or a change management gap. Clinical champions - doctors and nurses who translate frontline needs into product iteration - are essential to that process. "Trust in healthcare AI is not built by one successful demonstration. It is built through repeated, reliable performance in daily work," Gong said.
What separates the leaders in five years
Five years from now, Gong predicts the dividing line won't be who bought AI tools - many will have done that. The differentiator will be whether AI is embedded in the institution's operating system. Leading organizations will have strong, standardized data foundations, redesigned workflows built around human-machine collaboration, clear governance that maps tools to appropriate scenarios with defined oversight levels, and a culture that combines innovation with humility.
"The organisations that succeed will not be those that chase the most fashionable model, but those that continuously validate AI in real clinical settings and improve it with clinicians," she said. Success will be measured not by the number of algorithms deployed, but by whether care becomes more precise, efficient, equitable, and trusted.
For professionals working in healthcare technology, these principles align directly with the governance and workflow integration skills covered in AI for Healthcare training. Gong's three takeaways for HIMSS26 attendees reinforce the operational focus: treat trust as infrastructure rather than a soft issue, ground AI in evidence that is traceable and verifiable, and adapt solutions through local co-creation rather than exporting technology as-is.
Why this matters for healthcare professionals
Gong's framework translates into a practical checklist for any healthcare organization deploying AI. Before scaling a tool, teams need to verify that outputs are linked to clinical evidence, that the tool's role in the workflow is explicitly defined with matching oversight requirements, and that feedback loops exist to catch silent disengagement before it becomes entrenched resistance. If clinicians cannot explain why an AI made a recommendation, the system hasn't cleared the trust threshold - and won't move beyond pilot status, regardless of its technical accuracy.
Your membership also unlocks:
